I now analyse the relationship between employee turnover and the size of the employer and unemployment levels. Table 4.2 reports fixed-effects logistic hazard models for the probability of the worker moving to a new employer or into self-employment. The base category (zero) represents staying in the same firm. I control for a wide range of variables including job market experience, firm tenure, occupational level (skilled, semi-skilled or unskilled)10, full-time job, permanent contract, industry, quarter and
year-industry effects in all regressions. Firm tenure is the sum of the days that each worker has been working for the same employer, including all contracts that may have previously existed. Likewise, job market experience sums all working days that each individual has contributed to the Social Security system.
The baseline model in column 1 shows that the probability of job change declines monotonically with firm size. The negative coefficient of unemployment reflects that the hazard of turnover moves procyclically, meaning that the probability of employee
10
Skill levels are determined by the qualification requirements of the occupation defined by the Social Security contribution groups. Following García and Muñoz (2011) I classify the existing 10 contribution groups into: 1) skilled workers: engineers, graduates and chief and departmental heads; 2) semi-skilled workers: clerks, auxiliary workers and skilled labourers; and 3) unskilled workers: semi- skilled and unskilled labourers. These do not necessarily match with the educational attainment of individuals, since workers are many times overqualified for the job they perform. Indeed, this seems to be a common case in Spain, where the highest incidence of overeducated workers among OECD countries is found (Verhaest and Van der Velden,2010).
mobility declines with adverse labour market conditions. This might be a result of lower levels of hires in the economy overall and, therefore, more limited opportunities in the labour market. This would on one hand reduce the motivation of workers to resign and additionally, fewer workers would be able to find a job after being dismissed, although more separations would be expected to occur (Bruyere et al.,2011). Column 2 introduces interaction terms between unemployment and firm size categories and confirms the negative direct relation between firm size and employee mobility. Likewise, the effect of unemployment remains negative and bigger in magnitude, yet the positive coefficients of the interaction terms indicate that the negative effect of unemployment is attenuated and even reversed the smaller the size of the firm. Or to put it another way, the higher probability of turnover from the smallest firms, namely microfirms (i.e., with less than 10 employees), more than offsets the negative relationship between unemployment and job change. The overall effect of unemployment on the turnover hazard, however, remains negative for larger firms (i.e. with more than 10 employees). The rest of the included variables affect employee mobility in the expected direc- tions, except firm tenure. Despite the well documented evidence of higher probabilities of employee turnover at low tenure levels (Topel and Ward,1992;Farber,1999b) I find a positive effect of current firm tenure on employee mobility. I speculate that the ex- pected negative effect is probably captured by the highly positive effect of permanent jobs on turnover. As mentioned earlier, the Spanish dual labour market exacerbates the destruction of temporary jobs and leads to a greater rotation of workers with fixed- term contracts, which are by law shorter than 2 years11 (García-Serrano,1998). This
argument would also explain the higher turnover of unskilled employees, as it has been documented before (Güell and Petrongolo, 2007; García and Muñoz, 2011), exhibit
individual leaving their current employer, which I expect to vary as the economy slows and may operate differently in firms of different size. Thus, in the next model (column 3) I examine the probability to be dismissed, compared to the probability of staying in the same firm or leaving the firm voluntarily. This information is taken from the Social Security records as the contract termination reason provided by the employer as part of a compulsory administrative procedure. Employees in the smallest firm size category (i.e., microfirms, with less than 10 employees) are found to be more likely to be dismissed, although in this case the effect of firm size is not monotonically de- creasing. Workers in larger firms (i.e., with more than 10 employees) exhibit a lower probability to be dismissed than workers in largest firms (i.e., with more than 250 em- ployees). As expected, having a temporary contract explains a substantial proportion of the probability of an individual to be dismissed.
Another possible approach that I have examined to assess the impact of dismissals in the turnover probability is to include dismissals as an independent dummy variable in a model similar to the ones shown in columns 1 and 2. I present the results in table B.1 in the Appendix B. As expected, dismissals turn out to be the main factor explaining the probability to change the employer. Although this finding is interesting in that it demonstrates that turnover decisions are mostly involuntary it must be taken with caution, hence the reason to leave it for the appendix. My concern about including it in the main text arises because dismissals in the vast majority of cases, unless the job is seasonal and the individual is again hired by the same employer with no other job spell between them, necessarily imply changes in the employer. This means that introducing dismissals on the right- hand-side can constitute almost a “tautology” and make causal relationships difficult to infer.